Chinese President Xi Jinping used his 2026 New Year address to declare that 2025 marked a year when China’s technologies, including artificial intelligence and semiconductor chips, “reached new heights,” positioning the country as a leading rival to the United States in advanced computing. He praised Chinese artificial intelligence and chip firms for having “integrated science and technology deeply with industries, and made a stream of new innovations,” and said that “many large AI models have been competing in a race to the top, and breakthroughs have been achieved in the research and development of our own chips.” His remarks capped a year in which Chinese companies rapidly rolled out new frontier models and pushed to reduce dependence on foreign hardware, even as geopolitical tensions over semiconductors and data security intensified.
The year opened with Chinese artificial intelligence company DeepSeek launching its R1 model, which focused on complex tasks such as reasoning, coding and maths and at the time of its launch rivalled OpenAI’s o1, one of ChatGPT’s latest models. The launch of R1 sent US tech stocks plunging throughout the month of January, when American semiconductor chip giant Nvidia saw shares plummet by 17 per cent in one day, erasing $600 billion (€510 billion) in market capitalisation. DeepSeek followed up in September with DeepSeek-V3.2, an experimental version of DeepSeek V.31-Terminus that introduced a feature called Sparse Attention (DSA), which claims to reduce computing costs by 50 per cent without hurting performance, and the model can also generate large amounts of training data for artificial intelligence agents that operate without human supervision. According to descriptions on Hugging Face, DeepSeek said the model performs similarly to OpenAI’s ChatGPT-5, while its high-compute DeepSeek-V3.2-Speciale version outpaces GPT-5 and gets closer in reasoning to Gemini 3-Pro, although several European governments, including Italy, Denmark and the Czech Republic, moved to ban DeepSeek tools on official devices over security concerns.
Other Chinese tech giants advanced in parallel, with e-commerce group Alibaba releasing its Qwen2.5-Max artificial intelligence model in January and Qwen3-Max in September and claiming both outperformed the latest DeepSeek model and Anthropic’s Claude on some benchmarks, alongside a commitment from Alibaba to invest 380 billion yuan (€50.6 billion) into cloud computing and artificial intelligence in the next three years. Huawei announced new computing technologies and artificial intelligence chips to compete with Nvidia as Beijing and Washington clashed over access to Nvidia products, and in December President Donald Trump said the company could sell its H200 chips to approved customers in China in exchange for a 25 percent surcharge. At the same time, United States-based social media and artificial intelligence firm Meta moved to deepen its own presence in the space by agreeing in December to buy Manus, an artificial intelligence startup with Chinese roots, in a deal reported at $2 billion (€1.7 billion), following Manus’s launch of a general-purpose artificial intelligence agent for research and coding earlier in the year. Looking ahead, Xi said China will launch its “15th Five-Year Plan” in 2026 with recommendations that include “forward-looking plans” for future-facing sectors such as artificial intelligence, quantum technology and brain-computer interfaces, and Deloitte forecasts that during this period the proportion of Chinese funding for basic research may exceed 10% as more R&D is channelled into semiconductors and artificial intelligence, while China’s artificial intelligence chip market is expected to grow seven to nine times bigger than the 2025 market, which is worth up to $40 billion (€34 billion), outpacing global growth as the country builds domestic data centres, chips and computing networks.
